semi_industry_update - ookui/invest GitHub Wiki
Refer to: Semiconductor Industry
Reading Notes on Bernstein's report: "Global Semiconductors: New Playbook for a Maturing Industry"
- Semiconductor is matruing, growth rate is slowing;
- Consolidation trends take a twist. 巨头们继续高额投资（TSMC, Intel, Samsung），产业整合目前在只是在进行中。内存是整个产业整合比较好的，但是今年也遇到了PC/SP下滑导致的需求弱问题。
- SP在半导体行业中的收入在2013年起已经超过了PC，预计到2015年来自SP的Semi收入将近800亿美金（PC大概为500亿美金），因此这几年是SP拉动了整个Semi行业的增长；
- SP之后，没有看见新的增长点。向IoT/Wearable/Automobile等行业的Semi，不是量小，就是ASP低，有小机会，但是没有SP时代的大公司机会（如Qualcom, MediaTek, ARM...）。
来自SP的Semi收入目前已经占到了整个Semi的25%不到。
Source: Bernstein Report On Semi
Key points:
- 半导体行业增速放缓，以后consolidation是关键。DRAM, HHD, Foundry的整合已经完成的差不多了，接下去NAND Flash, Analog, Power等都有整合机会。由于整合，DRAM行业和HHD行业都经历了ASP提高，Margin提高的过程。
- 对于Memory: 由于密度 GB/Wafer上升的速度已经越来越慢，所以以后存储对wafer的需求会比较大。未来NADN Flash和DRAM都有可能被ReRAM所取代。
- IoT虽然增速很快，但是由于所使用的semi芯片ASP很低，对于整体半导体产业增速的贡献非常小。在2014年，PC使用半导体占全世界半导体收入的16%，smartphone占20%。
- 14-18的增速上看，Nonoptical Sensors, Optoelectronics, NAND以10.9%, 8.5%, 9.5%占据前三位。
- 中国进展：中国半导体的成长路径为 design-->OSAT-->foundry-->memory (先NADN Flash再DRAM）
- XMC和SMIC正在尝试进入NAND，可能和Spansion合作（Spansion的成长主要来自于MCU，flash业务并不好）；
- 中国财团并购了美国上市公司Integrated Silicon Solution (ISSI), 标志着中国进入DRAM的第一步。有可能是BOE与ISSI建立JV来共同发展DRAM
因此，最近将着重研究中国的半导体设计（如上海复旦，中国电子，同方国芯等）以及foundry（SMIC和HH）、半导体设备和材料公司的投资机会。
Updated: Oct 16, 2014
目前Microchip (MCHP US), ASML Holding (ASML NA), TSMC (2330 TT), Linear Technology (LLTC US)相继发布了3Q14的财报，试着从财报中看看当前ICT行业的趋势与状况。
TSMC从Samsung手中赢下iPhone 6的A8 代工订单 (20nm), 使得Samsung略过20nm投资，与Global Foundries一起强推14nm，期待2015年从TSMC手中赢回A9 (iPhone 7)订单。
据DB的数据，20nm和16/14nm比28nm的litho tool贵50%以上，并且复杂度更高。投资强度（Euro mn）：
20 & 16/14nm wafer需求（据DB在2014年9月的估计）：
- 2016年底要支持680mn high-end & mid-range Smartphones, 200mn high&mid-end tablets 需要
225K wafer/month; High-end non-SP products like PC CPU, FPGAs 需要另外的75K wafer/month. 而目前仅有TSMC的80K wspm。预计到2014年底，产能会到100-120K，这需要很大投资。
- ASML's expectation: global 20/16/14nm capacity to have reached 174 wspm based on 2014 ordering (available by mid-2015)
3Q14: sales is behind guide but due to a one tool push-out which should arrive 4Q. Order book is noticeably strong than expected at Euro1.4bn (jefferies est. E1.15bn). Q3 sales was largely driven by foundry customers (50% of total sales) versus 2Q when sales was largely driven by memory customers. One thing should be paid attention is that memory ordering accounts for 76% of 3Q14 order book. Is there over-investing in memory sector? 20/16/14nm business strongly depends on 10nm investment in 2015. It seems that Samsung has accelerated its 10nm process development in an attempt to gain competitive advantage . Technology: Samsung, Intel and TSMC spent money to support ASML to develop EUV (Extreme Ultraviolet Lithography machines). According to this article by Seeking Alpha contributor Confero: "ASML is nearly a monopoly company in the semiconductor lithography market with 74% of the market share."
Downstream:
-
Foundry: DB estimated Euro4.3bn of foundry litho CapEx in the pipeline for 2015/16.
由于20nm以及16/14nm制程的复杂性与高成本，以及下游客户需求的不确定性，使得1H14高制程投资减缓，1H14年ASML's revenue from foundry customers has substantially slowed from ~Euro500mn per quaarter to ~Euro230mn per quarter. 由于投资保守，所以20nm产能不足。2015年Apple，Qualcomm将会把20nm & 16/14nm单子释给Samsung/Global Foundry 和TSMC。
-
Memory: most its growth should be driven by upgrades of existing DRAM capacity. DRAM bit growth at mid 20% can typically be supported by technology conversions without adding new wafer capacity. ~500K wspm of capacity was estimated to be converted this year. New DRAM capacity is ~6X more expensive than conversion.
| company | Existing Capacity (2014E |
CAPEX | comments |
|---|---|---|---|
| Samsung |
~360K | Same level desipte of 50K wspm line 17 addition due to decrease in production in old fabs. 升级产能预计将从14年末的10K wspm逐渐增加到15年的40K wspm. |
|
| SK Hynix |
~260K | Expects to complete its new fab M14 by 2H15. Capacity is to be reduced to 230K wsqm due to relocation from M10 to M14. No increase in the capacity till 1H16. |
|
| Micron | ~350K | 2014E:2.8-3.2bn USD,2015E: 3.6-4.0bn USD (30% on NAND) |
No capacity expansion plan between 2H14~2015. Focus on completing its 25nm conversion and starting off with its 20nm ramp. |
- NAND: capacity growth is driven by technology conversions into higher notes and some capacity addition to sustain ~35%40% expected bit growth.
| Company | Existing Capacity (2014E) |
CAPEX | Comments |
|---|---|---|---|
| Samsung | ~330K | W4.2 trn (-10% yoy) |
No announced capacity addition plans but potential for capacity increase (e.g. Xi'an fab) |
| SK Hynix |
~150K | No disclosed capacity addition plans |
|
| Sandisk / Toshiba |
~470K | USD 1.4bn (859mn in 13) |
5% addition in 2H14, no other announced plan in 2015/16. ~50% share in iPhone 6 could drive upside |
| Micron | ~250K | 2014E: 2.8-3.2bn,2015E: 3.6-4bn (30% on NAND) |
No disclosed capacity addition plans |
台湾2013年半导体产业: MIC研發報告顯示，2013年台灣半導體產業產值達新台幣17,938億 元，相較於2012年成長幅度達16.1％，表現優於全球，各次產業的部分， 在 中低階智慧手持 的帶動之下，IC設計產業於應用處理器、面板驅 動IC等部分均有所斬獲，2013年台灣IC設計產值達新台幣4,549億元，相 較於2012年成長9.6％。IC製造產業部分，包括晶圓代工及記憶體產值達 新台幣9,697億元，與2012年相較成長幅度高達24.5％，其中主要原因是來自於 記憶體價格的復甦，帶動2013年記憶體產值大幅成長；此外，晶圓代工的部分，由於 台積電掌握28奈米先進製成領先優勢，亦對2013年台灣IC製造產業有所助益。IC封測產業亦有4.9％的穩定成長，產值預估達新台幣3,647億元。
Capex (in 3Q14 result conf call): 2014E: USD9.6bn in 2014, for 20nm expansion 2015E: 10bn+ USD, for 20nm expansion and 10nm engineering. 基本情况： China market: Demand for 4G SP & infrastructure in China remain robust in 4Q14. Inventory: TMSC expects inventory of fabless to be 2 days higher than the seasonal level by the end of 3Q14 and 2 days lower than the seasonal level by the end of 4Q14.